The Memory Decoding Challenge

A $100,000 prize for decoding a "non-trivial" memory from a preserved brain

By enabling dying people to have their brains and bodies preserved, we can provide them a reasonable chance of eventual revival by future medical advances. At least, that’s the hope that forms the foundational premise of brain preservation as a means of life extension.

Scepticism of this premise is common: How can a static, preserved brain possibly enable the eventual return of the dynamic neural activity that creates our thoughts and experiences? What if current preservation techniques, despite our best efforts, inadvertently destroy vital information about who we are? Given that neuroscientists still debate exactly how memories are stored and retrieved in living brains, how can we be confident they survive in preserved ones? Many conclude that revival has a negligible chance of ever becoming feasible - making preservation a costly pursuit of a false hope.

Given these conflicting positions, how should we evaluate the probability that preservation could actually work? A few approaches:

• First-principles analysis: We can examine each required step - from the initial preservation process through to eventual revival - and evaluate its likelihood of success. This means methodically asking questions like: What physical and chemical changes occur during preservation? What information needs to be retained? What technologies would be required for revival?

• Survey-of-experts approach: We can skip the difficulty of performing a comprehensive first-principles analysis, and instead defer to what neuroscientists and medical researchers think. For example, some colleagues and I recently surveyed the neuroscience community about preservation’s potential. We found the typical neuroscientist believes there’s a 40% probability that a well-preserved brain retains its long-term memories and could eventually be uploaded (preprint, blog post, news article).

• Experimental validation: While we can't yet test revival directly, we can design experiments that would strengthen or weaken our confidence in preservation's premises. For instance, if we could decode specific memories from a preserved brain, it would suggest crucial information survives the preservation process. Conversely, finding that key neural features are irretrievably lost - even with our best techniques - would raise serious doubts.

For an example of an experimental finding that strengthens the case for preservation, see the groundbreaking fruit fly brain model published in October of last year in Nature. A large collaboration of neuroscientists had recently mapped all 125,000 neurons and 50 million connections of the fly brain into a complete wiring diagram (a ‘connectome’). Using this data, the lead author Phillip Shiu and colleagues created a digital model of the fly brain where they could activate or silence any neuron and observe the cascading effects through the network.

The results were remarkable. When they activated sugar-sensing neurons in their virtual fly brain, it triggered a precise sequence of neural activity that ultimately activated motor neurons controlling the fly’s proboscis - essentially, making their digital fly stick out its tongue in response to detecting sugar. In the words of one of the project leads, Sebastian Seung: “Mind uploading has been science fiction, but now mind uploading — for a fly, at least — is becoming mainstream science.”

We shouldn't oversell this achievement though. The model had significant limitations:

• Each of the 125,000 neurons was simplified to a basic unit that just summed its inputs. Real neurons are far more complex, capable of significantly more sophisticated information processing.

• The model assumed neurons were silent unless stimulated, while real neurons have baseline activity.

• Only a few simple behaviours were tested.

• The model couldn't account for gap junctions, non-spiking neurons, internal states, or neuromodulators.

Despite these constraints, the fact that coherent behaviours could be read out at all from a static map of neural connections is encouraging for preservation advocates. When the researchers activated sugar-sensing neurons in their model, the resulting cascade of activity mirrored what we see in living flies with remarkable accuracy - over 90% of their predictions about which neurons would respond were experimentally verified.

Still, when we're thinking about what information needs to be retained to ensure our survival, eating reflexes are probably not at the top of the list. Most people would agree that preserving someone’s memories, personality, and other unique psychological properties is what really matters for maintaining their identity. While the feeding behaviours demonstrated in the fly study are a kind of ‘instinctual memory’, they're not specific to that particular fly’s experiences - they’re instead shared generically across all its fruit-loving fellows.

A more informative test of preservation’s premises would be to see whether a memory specific to that individual fly could be decoded from its connectome. Imagine if this fly had been raised eating vanilla-scented food, or if the vial in which it lived was roughly handled by a scientist wearing a jangly necklace. Could we detect a preference for vanilla over chocolate by examining its preserved brain? Could we somehow decode a learned fear of metallic sounds from its neural wiring?

The Aspirational Neuroscience community exists to encourage exactly these kinds of experiments.

The Memory Decoding Challenge

The Aspirational Neuroscience community's core initiative is the Memory Decoding Challenge: a $100,000 prize to be awarded to “the first group to decode a ‘non-trivial’ memory from a static map of synaptic connectivity.”

But what exactly counts as a ‘non-trivial’ memory? This was a key topic for the panel discussion at the Aspirational Neuroscience meetup in 2023 (recording). Some observable features of a brain or body are too simple to qualify as impressive examples of memory decoding - such as enlarged muscles indicating a history of frequent exercise, or scarring in neural tissue implying previous brain damage. These features tell us something about an animal's past, but they're ‘trivial’ in that they reveal little about its complex behaviours or internal experiences.

In contrast, examples of clearly non-trivial memories are:

• A zebra finch’s unique song: Each male zebra finch composes its own distinctive song during development - a sort of sonic signature. A research team could win the prize by examining a preserved bird’s auditory or motor cortex and accurately determining what its particular song had been.

• A rat’s maze navigation path: Behavioural neuroscientists often study learning by training rodents to navigate mazes for rewards. After multiple training sessions, they can measure learning by comparing the maze completion times of trained versus naive rats. Being able to examine a preserved rat’s hippocampus and decode which specific route it had taken would definitely count as decoding a ‘non-trivial memory’.

Of course, if a group managed to create a complete, high-fidelity brain emulation of an animal that perfectly replicated its learned behaviours, they would win the prize. But the challenge is deliberately designed to be achievable without requiring full brain emulation - that’s too high a bar for now.

Encouraging Progress Through Annual Awards

To encourage progress toward the main challenge, the Aspirational Neuroscience community also offers four $25,000 Annual Research Awards for papers that achieve milestones on the path to memory decoding¹. A few examples of previous awardees includes:

• Holler et al., 2020 - “Structure and function of a neocortical synapse” - (paper, journal club presentation)

  • Question: Can you determine how strongly one neuron influences another just by looking at the size of their synaptic connection?

  • Finding: The researchers demonstrated a linear relationship between synapse size and strength.

  • Significance: This suggests we can assign accurate connection weights in digital brain models purely by examining microscope images of preserved tissue.

• Choi et al., 2018 - “Synaptic correlates of associative fear memory in the lateral amygdala” - (paper, journal club presentation)

  • Question: Do collections of neurons storing the same memory (an engram) form stronger connections with each other compared to unrelated neurons?

  • Finding: Using a novel synapse-labelling technique, the team proved this was indeed the case.

  • Significance: This observable relationship between memory and connectivity suggests that strengthened connections are the physical substrate where memories are stored.

While these papers don't demonstrate actual memory decoding, they provide essential tools and foundational knowledge for future success.

Join the Community

If the prospect of memory decoding intrigues you, please consider:

• Nominating a paper for one of the Annual Research Awards.

  • Previous nominees are listed here.

• Joining the Aspirational Neuroscience Journal Club.

  • Discuss nominated papers and explore questions about brain mapping, modelling, and behavioural neuroscience.

  • Connect with others interested in these fundamental questions.

  • Sign up here!

  • Recordings of some of the previous presentations are available here.

• Attending the Aspirational Neuroscience meetup and Annual Award Ceremony

  • To be held around the Society for Neuroscience (SfN) meeting in San Diego, 15-19 November 2025.

  • Exact date and location to be determined.

The Aspirational Neuroscience prize money is only available due to a generous donation, while all other activities of the community are performed by volunteers. If you’re interested in contributing, we’d love to have you!

In the interest of transparency, you should know that the Aspirational Neuroscience community was founded by Ken Hayworth, who is also President of the Brain Preservation Foundation. The Memory Decoding Challenge is the spiritual successor to the Brain Preservation Prize, which was awarded in 2018 to the creators of aldehyde-stabilized cryopreservation, the first “[rigorously demonstrated technique] capable of inexpensively and completely preserving an entire human brain for long-term (>100 years) storage with such fidelity that the structure of every neuronal process and every synaptic connection remains intact and traceable using today’s electron microscopic (EM) imaging techniques.”

However, participating in the Aspirational Neuroscience community and its activities does not require you to endorse the feasibility or morality of brain preservation for life extension. The possibility of decoding memories from static maps of brain connectivity is broadly appealing to neuroscientists, and your expertise and insights are valued regardless of your position on preservation.

Indeed, a key strength of the Memory Decoding Challenge is its usefulness to both advocates and critics of preservation for life extension. Success would strengthen the case for preservation by providing clear proof that at least some important, person-defining memories remain intact within preserved brain structure. Conversely, if the prize conditions cannot be met despite prolonged and determined effort by the neuroscientific community, the case for preservation is weakened. This would suggest that static snapshots of the brain are insufficient to specify memories, and that this may only be possible by examining active, dynamic brain activity.

Brain preservation may be a viable method for stopping the terminally ill from dying and giving them a chance of placing their lifespan under their control. Or it might be a misguided and expensive pursuit of an impossible goal. The success or failure of the Memory Decoding Challenge will help resolve this uncertainty, and the Aspirational Neuroscience community would love your help finding out which it’ll be.

1. Despite the awards being ‘annual’, due to the COIVD19 pandemic and other factors, they’ve so far only been awarded twice: in 2019 and 2023. We’re trying to ensure this becomes more regular going forwards.